Named Entity Recognition Of Electronic Medical Records Based On Deep Learning

The combination of information technology and the medical field marks the advancement of traditional medicine to modern medicine.According to the time series,the electronic medical record records in detail the entire course of the patient’s disease development and clinical diagnosis and treatment during the hospital stay.In the application of electronic medical records,the application level of electronic medical records in all tertiary hospitals in our jurisdiction must reach the four-level level of grading evaluation.With the widespread application of electronic medical record systems,a large number of electronic medical records have been derived.How to use the textual information of medical records has become an urgent problem to be solved.Through the analysis of electronic medical record data,not only can the quality of clinical records of medical staff be effectively improved,but also the clinical decision-making level can be effectively improved through preset clinical practice rules,and thus the quality of medical care.Through the analysis of the characteristics of the electronic medical record text,the recognition of the named entity of the Chinese electronic medical record is carried out in-depth research from the application field,and the deep learning method is used for the recognition of the named entity of the electronic medical record.The following is the main research content of this article:(1)Construct a small-scale electronic medical record corpus.In this study,392 electronic medical records provided by a hospital were used to extract,sort,clean and label the original medical record data to construct a small-scale electronic medical record corpus for experiments.The self-built corpus and CCKS2018 public data set are used as the experimental data set of this article.(2)Use deep learning algorithms to conduct research on named entity recognition of electronic medical records.Analyze the text characteristics of electronic medical records,choose CRF as the baseline model,and conduct experiments on Bi-LSTM-CRF,IDCNN-CRF,Bert-Bi-LSTM-CRF,and Bert-IDCNN-CRF on two sets of data.Verification,using Word2 vec technology and Bert pre-training language model in feature representation.According to the experimental results,the results of Bert-Bi-LSTM-CRF are better than other models,so Bert-Bi-LSTM-CRF is selected as the basic model for subsequent experiments.(3)Propose an improved Bert-Bi-LSTM-CRF electronic medical record named entity recognition algorithm.The functions of the dictionary and the radical features of Chinese characters are respectively integrated into the character embedding as input to the model to enrich the semantic features.The experimental results show that,on the basis of the Bert-Bi-LSTM-CRF model,new features are added,and the F1 value of the electronic medical record named entity recognition is improved compared with the Bert-Bi-LSTM-CRF model on the two data sets.